Assistants don't know your menu
Without a direct source, AI assistants rely on scraped or outdated menus and can suggest items, prices or combos that don't exist.
Agentic Ordering
Agentic ordering for restaurants. Your menu, inside the assistant.
Guests are starting to ask AI assistants what to eat. menuOS lets them order your menu there, with accurate items, modifiers and availability for their chosen store, then hands the cart to your checkout so the customer stays yours.
Where it breaks today
Guests are asking. Assistants are guessing.
An order is more than a list of items
A real order depends on the store, the time of day, what's in stock and which modifiers are allowed. A general-purpose assistant can't see any of that.
Nobody wants to hand over the guest
Brands want the new channel without giving checkout, loyalty or first-party data to another intermediary.
How it works
Recommend, don't relinquish. Checkout stays with you.
Understand the request
The guest describes what they want in their own words. menuOS interprets it against your actual menu.
Build a valid cart
Items, sizes, modifiers and combos are resolved for the chosen store and daypart, including what's available right now.
Suggest within your rules
Recommendations and add-ons follow brand rules, so the assistant never offers what you wouldn't.
Hand off to your checkout
The finished cart opens on your website or app, where the guest pays, earns loyalty and stays your customer.
What's in the module
A new channel, on the data you already trust.
Agentic Ordering opens new channels such as AI assistant apps and messaging, and brings conversational ordering to the web and app channels you already run.
AI assistant apps
Put your menu inside assistants such as ChatGPT, built on open standards like the Model Context Protocol.
Menu-grounded answers
Every answer comes from your published menu, prices and availability, not from the open web.
Store-aware carts
Carts reflect the chosen location's menu, hours and 86s at the moment of ordering.
Brand guardrails
Control what the assistant suggests, how it describes items and which requests it declines.
Cart hand-off
A structured link carries the cart to your existing checkout, with attribution for the channel it came from.
Your own channels too
The same ordering capability can power conversational ordering in your own app and website.
Where it sits in menuOS
The experience layer. Built on everything below it.
Agentic Ordering reads the published menu from Menu Management, store details from Store Management and nutrition and allergen data from Product Lifecycle.
Because it sits on the same data as every other channel, what a guest orders in an assistant matches what the store can make, at the price the brand approved.
What you'll see move
Accurate orders, attributable revenue.
Order accuracy
Assistant-built carts that reach checkout without corrections.
Cart-to-checkout rate
Share of assistant-built carts that complete purchase.
Assistant-sourced orders
Orders attributed to AI channels, tracked like any other channel.
Time to launch
Time from kickoff to a live assistant experience.
Common questions
Frequently asked.
What is agentic ordering?
It's ordering carried out by an AI agent on the guest's behalf: the guest says what they want and the agent turns it into a valid order. menuOS provides the menu knowledge and ordering rules that keep those orders accurate.
Which AI assistants does menuOS support?
menuOS is open by default. It isn't tied to any one assistant, so it works with any agentic commerce framework or protocol your guests' assistants use, and extends to new ones as they emerge.
How do you stop the assistant making things up?
Answers are grounded in your published menu and checked against it before a cart is built. Items, modifiers or prices that aren't on your menu can't be added.
Do we need the other menuOS modules?
No. Agentic Ordering works on its own with your existing menu data, onboarded as part of setup. Menu Management is a natural next step, since it keeps every store's menu accurate at the source, but it isn't a requirement.